Jim Brain brain at jbrain.com
Fri Jan 2 21:51:07 CST 2009

Tony Duell wrote:
>> Tony Duell wrote:
> If oyou use one switch to select between SS/80 and Amigo (protocol), that 
> leaves 2 switches for drive type. Not really enough. And what are you 
> going to do about sleecting the HPIB primary address (HP units normally 
> had 3 switches for this, allowing addresses from 0 to 7 only).
Not sure.  I do have the ability to add an entire latch of switches on 
the IDE bus (like I am adding the IEEE-488 support.)  On the current 
unit, you send commands to the unit through the command channel to 
change such things, so maybe that is another option (pick the command 
set from the front, then issue a command to pick a device type and 
choose an alternte address).  On the latter, the uIEC supports the same 
idea.  It boots as device #10, and you can modify it via software.  So, 
you put it alone on the bus, issue a command to change the number, saves 
it in EEPROM, and then you plug in the rest of the units.

> Are those still being made? I wouldn't want to design round a chip that's 
> in short supply.
Actually, they are.

When I was composing the email, I initially wrote about using latches 
and transceivers, but then decided (with this crowd) I should do a bit 
more investigation before I posted.  Along the say, I found the 75XXX 
units, and they weren't that bad in price.

But, right after I rewrote the msg and sent it, I realized that I'd 
still need 2 latches to hold the outputs for the 16 bits of IEEE data, 
and another latch to save off the TE state (on the 75XXX ICs).  As well, 
I'd be locked into IEEE 488 usage.  So, I scrapped my idea, and went 
back to my original ideas, TPIC6B273s and 74LS245s.  It's the same 
price, and I could support more interfaces.
> FWIW, I've had some early HP HPIB units on the bench recently, and HP 
> used '38s as driers (open collector) and '14s as receivers (schmitt 
> trigger inputs). Of course using thsoe put the package count up.
'38s are somewhat iffy, in my mind.  MC3446 and the 75XXX line sink 
100mA, and the '38 is 48, max.  Coupled with the higher parts count (4 
'38s, plus 2 '374s = 6 parts as opposed to 2), the price is only 
marginally better (3.40 versus 2.40).

If the solution finds value outside of the vintage community, I could 
see one on a long GPIB chain, and I'd hate to fail in that case.  With 
150mA sink, I'll be the last one on the bus to fail.


More information about the cctalk mailing list